Background: Although ethics consultation has been introduced to clinical practice for many years, the results of empirical studies to evaluate the effectiveness of ethics consultation are still controversial. The design of randomised controlled trials is considered the best research design to evaluate the effect of a clinical practice on the outcomes of interests. In order to understand the effects of ethics consultation, we conducted this search for studies with the design of randomised controlled trials to evaluate ethics consultation.
Objective: To provide an integrated review of studies with the design of randomised controlled trials to evaluate the effectiveness of ethics consultation.
Methods: PubMed was used to search for studies using the randomised controlled trial design to evaluate the effectiveness of ethics consultation. The search term used was “ethics consultation”. The selection criterion was limited to “randomised controlled trial”.
Results: Four articles that met both search criteria were retrieved. One of these articles reported a study that did not actually use the design of a randomised controlled trial and is excluded from the following discussion.
Conclusions: To apply randomised controlled trials to evaluate the effectiveness of ethics consultation is extremely difficult as long as two issues are not resolved: the standardisation of ethics consultation and a placebo for ethics consultation to eliminate the placebo effect. Thus, the results generated by the design of randomised controlled trials are always problematic. Furthermore, as long as the two issues exist, the results generated by the design of quantitative research methods always pose problems.
Statistics from Altmetric.com
Ethics consultation has been introduced to clinical practice for decades. It is a service provided by an individual consultant, a group of skilled people or a committee to help patients, surrogate decision-makers, families, healthcare providers or other involved moral stakeholders to address the conflict or the uncertainty of value-laden issues in clinical practice.1 Empirical research designs are usually used in evaluating a clinical service. These studies can be conducted by the use of qualitative methods, quantitative methods or both. Studies in evaluating ethics consultation were first carried out in the 1980s.2 Until now, most evaluative studies have used quantitative methods. The design of a randomised controlled trial (RCT) is the best way to minimise bias in order to ascertain intervention effects. The RCT is considered the gold standard research design for evaluating the effectiveness of a new clinical intervention and also provides a direct test of a causal hypothesis.3–5 This article begins by providing an integrated review of RCTs in evaluating ethics consultation. Our purpose is not to criticise each individual study, but to address the difficulties in using the RCT design to evaluate the effectiveness of ethics consultation.
To retrieve articles reporting the use of an RCT to evaluate the effectiveness of ethics consultation, we used the huge database PubMed. The search term used was “ethics consultation”. The selection criterion was limited to “randomised controlled trial”. Four articles that met both search criteria were retrieved. One of these articles reported a study that was not actually an RCT and is excluded from the following discussion. The other three articles are reviewed in chronological order, particularly focusing on their design and results.
RANDOMISED CONTROLLED TRIALS IN EVALUATING ETHICS CONSULTATION
The first study that used an RCT to evaluate the effectiveness of ethics consultation was published in 2000.6 In this study, 70 patients were successfully recruited from medical and paediatric intensive care units (ICUs). Patients with value-laden treatment conflicts were identified by experienced nurses and randomly assigned to the intervention group (ethics consultation offered) and the control group (ethics consultation not offered). Crossovers between the two groups were allowed. Twelve patients in the intervention group did not complete ethics consultations and three patients in the control group requested and received ethics consultations. The analyses of this study were based on the intention-to-treat principle, which states that the primary analysis of data in an RCT should compare subjects based on the group to which they are randomised, regardless of their crossovers.7 Patients were evenly divided into the two groups. The control and intervention patients were similarly matched with regard to demographics, surrogate decision-makers, principal diagnosis and medical insurance payers. The researchers reported that the mortality in patients who received ethics consultations did not increase (p = 1.00). Among patients who did not survive to hospital discharge (21 controls and 21 interventions), those who received ethics consultations tended to have a shorter ICU stay (p = 0.03), a reduction in days receiving artificial nutrition and hydration (p = 0.05) and a reduction in days receiving artificial ventilation (p = 0.05).
In the second article, a larger, multisite RCT study was performed in adult ICUs of seven hospitals.8 Five hundred fifty-one patients with value-related treatment conflicts arising during their ICU stays were identified by experienced nurses. Among those patients, 278 were randomised to the intervention group and 273 were randomly assigned to the usual-care group. Only 211 of the 278 participants assigned to the ethics consultation arm actually completed ethics consultations, and 196 of the 273 patients assigned to the usual-care arm did not receive ethics consultations. Data were analysed according to the intention-to-treat principle. Confounders such as patient demographics, diagnosis, surrogate decision-makers and medical insurance payers were not significantly different between the two groups. Mortality in the two groups did not differ significantly. Among those patients who did not survive to hospital discharge, the intervention group had significantly shorter hospital stays (p = 0.01), ICU stays (p = 0.03) and days receiving mechanical ventilation (p = 0.03). In contrast, among patients who were discharged alive, there was no difference between the two groups in relation to hospital stays, ICU stays and days receiving mechanical ventilation.
In the third article, researchers estimated the costs of hospital stays for patients who did not survive to hospital discharge, using the dataset collected from the second study.9 Patients were randomly assigned to the intervention group or the usual-care group. The intention-to-treat principle was followed in data analysis. In this dataset, 499 of the total 551 patients were included in the cost analysis. There were no significant differences between the two groups in age, gender, ethnicity, diagnosis or survival. Among those who died in hospital, the intervention group (n = 156) had significantly shorter hospital stays (p = 0.016) and lower hospital costs (p = 0.021) than the usual-care group (n = 144). In contrast, there were no significant differences in hospital stays or costs between the intervention patients and the usual-care patients who survived to hospital discharge.
This review shows similar findings in the three studies. For patients who did not survive to hospital discharge, ethics consultations were significantly associated with shorter ICU stays, shorter hospital stays, less use of life-sustaining treatments and lower hospital costs. In contrast, no significant differences were identified in the four variables for patients who survived to hospital discharge. It seems that the intervention of ethics consultation was beneficial to patients who did not survive to hospital discharge and was not harmful to patients who did survive. In order to better understand these study results and to discuss the significance of the findings, further examination of the use of the RCT design in evaluating the effectiveness of ethics consultation is necessary.
A causal relationship can be tested if there are two groups of subjects that are similar or equivalent in all relevant characteristics and the putative causal factor is applied to one group. It can be supported if the frequency of defined outcomes is significantly higher or lower in the group with the putative causal factor. The best study design of this sort is an RCT. The assignment of intervention for each subject is achieved by a random procedure. Because the subjects receiving or not receiving the intervention being tested are selected randomly, the two groups are theoretically similar to each other with regard to all relevant characteristics. Thus, the differences in outcomes can be directly attributable to the difference in the intervention. In addition to randomisation, controlling bias arising from subjects and observers is the other key to a successful RCT. As Elwood pointed out, “The most important sources of bias are variation in the subject’s response and to the method of assignment, and variation in the observer’s response.”3 In order to best ensure that neither subjects nor observers are aware of which group they are in, a method called “double blinding” can be used. For example, a researcher plans to conduct an RCT to evaluate the effectiveness of a new drug. To ensure that patients and physicians are unaware of which drug patients are receiving, the new drug must be made up to smell, look and taste like the conventional one. Patients and physicians are blinded to which drug patients receive in the study. This design avoids bias from both patients and physicians. If either knows that they are in the group using the new drug, psychological factors are highly likely to elicit improved responses.
In the hierarchy of research designs, the results of RCT provide the most rigorous evidence to support the causal relationship between the independent variable (the intervention) and the dependent variable of interest (the outcomes).3 4 However, for practical reasons this research design is not simple to carry out.
In the three articles considered above, all the studies were performed by randomly assigning subjects to the intervention group (ethics consultation offered) and the control group (ethics consultation not offered). The major strength of these studies is that data were analysed based on the intention-to-treat principle. When ethics consultation is effective but crossovers are substantial, following the intention-to-treat principle to analyse data tends to underestimate the magnitude of the effect of ethics consultation that will occur in adherent patients.7 That is, if there had been no crossovers in the three studies, the true effect of ethics consultation should have been much greater than was measured. Here, however, we argue that the use of the RCT design in evaluating the effectiveness of ethics consultation is inevitably and methodologically problematic.
The first methodological concern with regard to using an RCT to evaluate the effectiveness of ethics consultation is that ethics consultation is not standardised. RCTs have been used most extensively in evaluation of treatments in clinical medicine, that is, the comparison of a new drug with a conventional one. A crucial characteristic of the two drugs being compared is the standardisation of both. That is, first, the composition of a drug does not vary from tablet to tablet. Second, the pharmacological and physiological interactions between a drug and a patient’s body are usually identical as long as a patient is eligible to enter the trial. In contrast, as Craig and May suggested, “At a practical level, what ethics consultation is meant to do will itself differ from case to case.”10 Ethics consultation is not standardised in three aspects: structure, process and the case that an ethics consultant is faced with.
Structure of ethics consultation refers to the organisation that provides the consultation service and the qualifications of the consultants, such as their educational and training background and their personality traits.11 Each organisation has its own policy about ethics consultation. Its guidelines may not completely conform to the general goals of ethics consultation. The guidelines of a university-affiliated medical centre may be different from those of a community hospital. In addition, ethics consultants with different educational and training backgrounds may execute ethics consultation differently.
The process of ethics consultation is defined by Fox as “the interactions between the intervention and individuals served by it”.11 There is no rule with regard to how the interactions between an ethics consultant and other participants should be. An ethics consultant may decide to facilitate consensus among involved parties (the pure facilitation approach) in one case, or to be the primary moral decision-maker (the authoritarian approach) in another.12 These different approaches, certainly, will lead to different processes of ethics consultation. Furthermore, ethics consultation can be performed by an individual consultant, a team or a hospital ethics committee.1 Ethics consultations performed by different parties may also result in different processes.
Also, what ethics consultants will do depends on the particular case that they are faced with. For example, an ethics consultant may have been consulted by an attending physician to assist with a patient’s end-of-life decision-making in a medical ICU yesterday, been contacted by a nurse just a minute ago to resolve care plan conflicts between an attending physician and the nurse in a general ward, and have a scheduled a meeting tomorrow with a paediatrician and a newborn baby’s parents to discuss medical futility in a neonatal ICU. Ethics consultants are faced with greatly differing cases. Not surprisingly, such complexities in structure, process and a myriad of unique cases threaten the standardisation of ethics consultation. Accordingly, using an RCT to evaluate the effectiveness of ethics consultation, which is not standardised, will always entail problems. The results generated from the RCT are therefore questionable.
Our second methodological concern is double blinding. One of the crucial factors of success in an RCT is to blind the subjects and the observers. The best way to blind them is to use a placebo—a fake intervention, or an inert substance that smells, looks and tastes like the active agent being evaluated but does not have any inherent power to result in outcomes of interest.4 A placebo is considered inert because it produces only the placebo effect. The placebo effect is the psychological feeling of improvement in outcomes not attributable to the intervention. Many theories have been offered to account for this effect.13 14 It is believed that a subject’s or an observer’s beliefs and hopes about an intervention may implicitly affect the outcomes.14 The purpose of using a placebo in a study is to eliminate the placebo effect. If the control group is not given a placebo, researchers cannot distinguish whether differences in outcomes between the intervention group and the control group are attributable to the intervention or to a pure placebo effect. By use of a placebo, subjects and observers cannot know which group they are assigned to, thus reducing the placebo effect. This explains why double blinding is strictly required to make an RCT successful.
As Fox and Tulsky pointed out, “Some may argue, for example, that ethics consultation is so obviously beneficial that formal evaluation is a waste of time”, and, “We, like many others in the field, firmly believe that ethics consultation is beneficial.”15 It is widely accepted by people in the field of clinical medicine that ethics consultation is beneficial. They believe that ethics consultation helps promote each moral stakeholder’s rights and quality of care. These beliefs and hopes with regard to ethics consultation lead to difficulties in the research design in evaluating the effectiveness of ethics consultation. As discussed above, those three RCT studies examined the effectiveness of ethics consultation. There is no doubt that most of the physicians, nurses, patients or surrogates, and other moral stakeholders in those studies may believe and hope that ethics consultation is beneficial. In the intervention group, people with these beliefs and hopes participated in ethics consultations. It is impossible to rule out the possibility that their behaviour was implicitly influenced by their beliefs and hopes. Therefore, the outcomes of the intervention group are the results of the true effect of ethics consultations plus the psychological feelings of improvement. In contrast, participants in the control group, who also may believe and hope that ethics consultation is beneficial, did not receive a placebo that was indistinguishable from an ethics consultation. Their behaviour was not changed or influenced by their beliefs and hopes. Thus, the outcomes of the control group are the results of no psychological feelings of improvement plus no ethics consultation. As a result, to compare the two groups, we cannot tell whether the differences in outcomes are attributable to the true effect of ethics consultation or the beliefs and hopes that ethics consultation is beneficial.
One ideal way to reduce the placebo effect of the three RCT studies is to introduce a placebo in the control group. Of course, participants in the control group also may believe and hope that ethics consultation is beneficial. If a placebo that is indistinguishable from an ethics consultation is given to the control group, participants in this study won’t know which group they are assigned to. All subjects and observers are blinded to the introduction of a placebo to the control group. The placebo effect in the study will be eliminated. Thus, the differences in the outcomes can be attributable to the difference in ethics consultations. However, is a placebo of this sort possible? The placebo must be perceived as being like an ethics consultation but must not have any inherent power that an ethics consultation has in resolving ethical conflicts and dilemmas. Indeed, a placebo that is indistinguishable from ethics consultation, we believe, does not exist to date. Thus, it is extremely difficult to examine the effectiveness of ethics consultation correctly by using an RCT.
We argue that the use of an RCT to evaluate the effectiveness of ethics consultation is problematic. Is there any other quantitative research design that provides scientifically convincing evidence, though not as rigorous as evidence provided by the design of RCT, to support the causal relationship? As pointed out by Campbell, the minimum of generalisable scientific evidence must have at least one comparison group.16 Without a comparison group, little can be concluded about the effect of the intervention. Accordingly, a single-group descriptive study cannot provide generalisable scientific evidence to support the effectiveness of ethics consultation: at least one comparison group is required for studies to evaluate the effectiveness of ethics consultation.
There are different methods of doing group-comparison studies, such as individual matching, group matching or multivariate logistic regression. For individual matching, all the possible confounders between the intervention and the outcomes are matched based on an individual subject. For example, if the control group contains a subject who is a 45-year-old African–American female high school graduate, married, with two daughters and one son, a subject in the intervention group should be matched by all these characteristics. For group matching, matching is based on group-level characteristics. For example, 30% of the subjects in the control group are male, the mean age is 59.5 years and 25% of them are African–Americans. An intervention group should be matched on the basis of those group-level characteristics. After individual or group matching, the difference in outcomes can be directly attributable to the difference in the intervention. The other method of doing a group comparison study is multivariate logistic regression. After controlling confounders, the association between the intervention and the outcome can be examined. However, regardless of how sophisticated investigators are in designing group-comparison studies, they will inevitably encounter the two difficulties that we have discussed above: the impossibility of standardising ethics consultation and the unavailability of an appropriate placebo to eliminate the placebo effect. As long as these two difficulties are not resolved, evaluating the effectiveness of ethics consultation by group-comparison studies will always encounter problems. Because a single-group descriptive study cannot provide scientifically generalisable evidence in evaluation of ethics consultation, and because the lack of standardisation and of an appropriate placebo group makes group-comparison studies unachievable, evaluating the effectiveness of ethics consultation by quantitative research methods is extremely difficult.
Many articles have raised concerns about evaluating the effectiveness of ethics consultation.10 17 18 Although an RCT design has been used to evaluate the effectiveness of ethics consultation and has provided more scientifically convincing evidence than any previous studies, we argue that using an RCT to evaluate the effectiveness of ethics consultation is extremely difficult and always encounters problems, namely, the lack of a placebo for ethics consultation to eliminate the placebo effect and the lack of standardisation of ethics consultation. Results generated from RCTs are therefore questionable. Moreover, as long as these two issues still exist, evaluating the effectiveness of ethics consultation by quantitative research methods will remain problematic.
The authors want to thank Professor Stuart Youngner for his thoughtful comments on the early draft of this paper.
Funding: This is a self-funded paper.
Competing interests: None declared.